Bright shiney new Training Infographic question

um

Jul 2019

5:20pm, 2 Jul 2019

1,087 posts

0

um

Disabling GC altitude correction changes it to 1132ft ...

( and it was primarily a flat canal run, although bit of a downhill to start and uphill to finish, but not 1132ft! )

oi you

Jul 2019

5:46pm, 2 Jul 2019

437 posts

0

oi you

The way GPSs work is very clever, and I'm not but my best guess is that (ignoring the altimeter information, which is what I believe happens, on some websites anyway,) the GPS gadgets estimate their current location by reference to very accurate time signals from the satelites. It then calculates distance, pace etc from the series of position fixes.
The fixes have an accuracy limit, and the map info on the gadget used to calculate distances can't be as detailed as FE might use.
They're good, but not that good.

larkim

Jul 2019

6:11pm, 2 Jul 2019

7,976 posts

0

larkim

Yes, but most gadgets don’t have maps at all - all they do is record the positions. The maps that are displayed on clever devices (some watches, most phones) etc are just a nice overlay - they don’t relate to the position being recorded by the device in terms of applying any correction to where they are. The exception to this is satnavs for cars, where the software cleverly “snaps” any error to match the road, so it doesn’t assume you are driving off road, or on the wrong side of the road or the pavement etc.

You’re right that each individual position recording is prone to error, and that’s where the clever error correction on watches or other devices comes into play to smooth out the route a little. I don’t know precisely how it works, but I’d imagine they look at a small series of points recorded and create a track which is more consistent. Otherwise, as every point recorded could be 3-4m away from where it “really” is you’d end up with a trace which was very jaggedy - as some early implementations of GPSr devices recorded.

My eldest is studying geomatics at Uni, so whilst I’m not an expert by any stretch of the imagination, I’ve made sure I’m correct insofar as his technical understanding of these things!!

Badger

Jul 2019

6:25pm, 2 Jul 2019

11,000 posts

0

Badger

oi you: The location calculation is about how you put it, yes, but there's an extra factor in most systems, where they use the Doppler shift of the signals to estimate your speed, and use filtering of all that information to provide the regular position updates with minimal wildly wrong points.

The horizontal location is pretty precise and consistent most of the time. Vertical accuracy based on GPS alone is considerably poorer, maybe 3x with a good sky view, much worse when only a few satellites are visible.

All units are calculating distance horizontally on the fly, and the filtering and some other maths I won't go into unless anyone asks means that the errors in the distance don't really accumulate significantly, so GPS has to be having a pretty bad day to be more than 1% off.

There are no maps in most units, and other than some cycling units, none of them lock to paths or roads, so mapping is irrelevant. FE is not using mapping information to change horizontal distance anyway.

As far as vertical distance goes: anything with a barometric altimeter will be measuring cumulative ascent and descent as it goes, and it'll be quite accurate if the weather is stable. A few very new Garmins have DEM (digital elevation mapping) to calibrate the altimeter continuously, and when that's ironed out it should be very accurate (I've heard of some issues so far, though).

Units without a barometer sometimes do calculate ascent/descent along with current elevation from GPS info, sometimes don't, and it's often pretty noisy, hence the mapping calculations a lot of websites apply. There are no devices on the market with DEM for elevation but no barometer.

If you have a Suunto Ambit or similar, or a Garmin Fenix or high-end Forerunner, it's highly likely that it will measure ascent and descent more accurately than at least some websites. The reason is that it's giving you a true measure of elevation at every second; the elevation maps really aren't that precise; many of them are only about 90m squares, some of them have smoothing and interpolation and some don't, some of them have data from pole to pole and some don't (not an issue unless you're at least up at the north end of Norway IMS). Sometimes a small error in location can turn into a huge error in mapped elevation - say a path along a cliff top, where a few metres to one side would have you at sea level (literally if you're not careful

) Bridges can confuse elevation mapping - your horizontal location doesn't say whether you're going over or under.

FE has my race on Saturday at 1873m ascent, 1874 descent; the watch says 1984 gain, 1923 loss (which reflects the weather worsening through the day). I can live with a difference like that. On the other hand, a mostly-clifftop run a couple of weeks earlier said 359m on the watch, 438m on the site I planned it on, and 491m on FE, and I think that's the mapping resolution plus wandering GPS on steep slopes.

oi you

Jul 2019

8:52pm, 2 Jul 2019

438 posts

0

oi you

See! I said it was clever stuff. Thank you for the better explanations.
As the old joke has it - I'm none the wiser, but I am better informed.

Tonights run - 200(approx) laps round (and round) a cricket field as flat as you'd like. Total ascent 57m. I think not.

jabberknit

Jul 2019

1:21pm, 3 Jul 2019

2,907 posts

0

jabberknit

All very educational, people, thank you very much, but is there anything I can do to get my Fetch ascent figures anywhere nearer my Garmin Connect / ~~strava~~ ones?

um

Jul 2019

4:17pm, 3 Jul 2019

1,089 posts

0

um

I suspect entering a feature request in the feature list suggesting the Fetch elevation may sometimes be suspect may be enough to trigger attention?

Got something to say?

About This Thread

Related Threads

Free training & racing tools for runners, cyclists, swimmers & walkers.

um Jul 2019 5:20pm, 2 Jul 2019 1,087 posts	0 um Disabling GC altitude correction changes it to 1132ft ... ( and it was primarily a flat canal run, although bit of a downhill to start and uphill to finish, but not 1132ft! )
oi you Jul 2019 5:46pm, 2 Jul 2019 437 posts	0 oi you The way GPSs work is very clever, and I'm not but my best guess is that (ignoring the altimeter information, which is what I believe happens, on some websites anyway,) the GPS gadgets estimate their current location by reference to very accurate time signals from the satelites. It then calculates distance, pace etc from the series of position fixes. The fixes have an accuracy limit, and the map info on the gadget used to calculate distances can't be as detailed as FE might use. They're good, but not that good.
larkim Jul 2019 6:11pm, 2 Jul 2019 7,976 posts	0 larkim Yes, but most gadgets don’t have maps at all - all they do is record the positions. The maps that are displayed on clever devices (some watches, most phones) etc are just a nice overlay - they don’t relate to the position being recorded by the device in terms of applying any correction to where they are. The exception to this is satnavs for cars, where the software cleverly “snaps” any error to match the road, so it doesn’t assume you are driving off road, or on the wrong side of the road or the pavement etc. You’re right that each individual position recording is prone to error, and that’s where the clever error correction on watches or other devices comes into play to smooth out the route a little. I don’t know precisely how it works, but I’d imagine they look at a small series of points recorded and create a track which is more consistent. Otherwise, as every point recorded could be 3-4m away from where it “really” is you’d end up with a trace which was very jaggedy - as some early implementations of GPSr devices recorded. My eldest is studying geomatics at Uni, so whilst I’m not an expert by any stretch of the imagination, I’ve made sure I’m correct insofar as his technical understanding of these things!!
Badger Jul 2019 6:25pm, 2 Jul 2019 11,000 posts	0 Badger oi you: The location calculation is about how you put it, yes, but there's an extra factor in most systems, where they use the Doppler shift of the signals to estimate your speed, and use filtering of all that information to provide the regular position updates with minimal wildly wrong points. The horizontal location is pretty precise and consistent most of the time. Vertical accuracy based on GPS alone is considerably poorer, maybe 3x with a good sky view, much worse when only a few satellites are visible. All units are calculating distance horizontally on the fly, and the filtering and some other maths I won't go into unless anyone asks means that the errors in the distance don't really accumulate significantly, so GPS has to be having a pretty bad day to be more than 1% off. There are no maps in most units, and other than some cycling units, none of them lock to paths or roads, so mapping is irrelevant. FE is not using mapping information to change horizontal distance anyway. As far as vertical distance goes: anything with a barometric altimeter will be measuring cumulative ascent and descent as it goes, and it'll be quite accurate if the weather is stable. A few very new Garmins have DEM (digital elevation mapping) to calibrate the altimeter continuously, and when that's ironed out it should be very accurate (I've heard of some issues so far, though). Units without a barometer sometimes do calculate ascent/descent along with current elevation from GPS info, sometimes don't, and it's often pretty noisy, hence the mapping calculations a lot of websites apply. There are no devices on the market with DEM for elevation but no barometer. If you have a Suunto Ambit or similar, or a Garmin Fenix or high-end Forerunner, it's highly likely that it will measure ascent and descent more accurately than at least some websites. The reason is that it's giving you a true measure of elevation at every second; the elevation maps really aren't that precise; many of them are only about 90m squares, some of them have smoothing and interpolation and some don't, some of them have data from pole to pole and some don't (not an issue unless you're at least up at the north end of Norway IMS). Sometimes a small error in location can turn into a huge error in mapped elevation - say a path along a cliff top, where a few metres to one side would have you at sea level (literally if you're not careful ) Bridges can confuse elevation mapping - your horizontal location doesn't say whether you're going over or under. FE has my race on Saturday at 1873m ascent, 1874 descent; the watch says 1984 gain, 1923 loss (which reflects the weather worsening through the day). I can live with a difference like that. On the other hand, a mostly-clifftop run a couple of weeks earlier said 359m on the watch, 438m on the site I planned it on, and 491m on FE, and I think that's the mapping resolution plus wandering GPS on steep slopes.
oi you Jul 2019 8:52pm, 2 Jul 2019 438 posts	0 oi you See! I said it was clever stuff. Thank you for the better explanations. As the old joke has it - I'm none the wiser, but I am better informed. Tonights run - 200(approx) laps round (and round) a cricket field as flat as you'd like. Total ascent 57m. I think not.
jabberknit Jul 2019 1:21pm, 3 Jul 2019 2,907 posts	0 jabberknit All very educational, people, thank you very much, but is there anything I can do to get my Fetch ascent figures anywhere nearer my Garmin Connect / ~~strava~~ ones?
um Jul 2019 4:17pm, 3 Jul 2019 1,089 posts	0 um I suspect entering a feature request in the feature list suggesting the Fetch elevation may sometimes be suspect may be enough to trigger attention?